Prabal Mukherjee

Deep Analysis on Proteins Interaction Using Graph Neural Networks (GNNs)

Author and Collaborator: Arumay Pal, Post doctoral researcher, Weizmann Institute of Science

Challenge:

• This research involves leveraging GNNs for graph-level or subgraph-level prediction models to analyze and predict the feature qualities between proteins.
• A critical first step in this project is constructing a high-quality dataset that encapsulates unbiased and balanced training samples. Building such a dataset requires an effective preprocessing pipeline and a thorough feature analysis process. Following preprocessed pipeline features will be predicted using deep learning methodology using Graph Neural Network.
• The research will leverage pytorch, pytorch-geometric in high performance GPU based training and testing.

Challenge: Develop robust solutions to support and enhance a proprietary CRM ecosystem for complex business needs.

Process:

• Utilized C#, Python, PySpark, ML, Angular, ECS OpenShift, Elastic ELK, APM, and CI/CD to develop tailored solutions.
• Leveraged PySpark for big data ETL processes and analysis of massive unstructured call reports for CRM contacts and accounts.
• Designed and implemented machine learning models and data processing pipelines. And developed frontend solutions using Angular that is integrated with backend micro services, deploying on ECS OpenShift for scalability and high availability.
• Monitored performance using Elastic ELK and APM.
• Collaborated with different teams, mentored junior developers and interns, and managed enterprise excellence for the overall CRM system.

Result:

• Delivered scalable solutions enhancing CRM capabilities, improving sales team productivity through advanced analytics and automation.
• Achieved a more reliable and efficient CRM system, leading to increased user satisfaction and operational efficiency.

Challenge:Design and develop critical risk management applications and data solutions to support trading and risk assessment for MBS/SP operations.

Process:

• Led the design and development of the Market Risk application (Titan), Credit Risk Applications ETL workflow for Credit Insight, and Middleware for the Longevity Trade booking app.
• Employed a tech stack including C#, ASP.NET, TIBCO EMS, SQL, Python, ML, PySpark, SparkNLP, and Databricks to build robust and efficient solutions.
• Utilized Databricks and PySpark for big data exploratory data analysis (EDA) and data science solutions, providing quarterly reviews for equity analysts of large US companies by leveraging alternative data.
• Led and mentored small development teams, collaborating with cross-functional teams for various firm-wide programs.

Result:

• Successfully delivered high-performance risk management applications, improving trading and risk assessment capabilities.
• Enhanced the efficiency and accuracy of risk evaluations through advanced data analysis and machine learning solutions.
• Streamlined risk data workflows, leading to better decision-making and operational efficiency for equity analysts.

Challenge: Replaced a decade-old vendor product with a new, custom-built event management system.

Process:

• Led development of Event management system from the ground up using ASP.NET MVC, C#, and SQL Server.
• Architected the system to meet modern requirements and improve functionality and user experience.

Result:

• Successfully delivered a robust event management system that enhanced operational efficiency and user satisfaction.

Challenge: Develop a real-time pre-trade booking system for all MBS desks that integrates with sales and tracks activity for each security.

Process:

• Designed, developed, and implemented pre-trading system single-handedly using C#, Sybase, TIBCO EMS, and Infragistics.
• Integrated system with sales to maintain activity records, known as "colors" for MBS traders.

Result:

• Delivered a robust pre-trade booking system that improved the efficiency and accuracy of pre-trade activities for MBS desks.

Challenge: Develop and support a real-time bond trading system for Smith Barney's retail business, integrating with external street counterparties and live market data.

Process:

• Developed and supported CMAT system using C# 3.5, WinForms, Infragistics 8.3, and Java MSGQ.
• Integrated the system with external counterparties like ValueBond, BondDesk, and MuniCenter and connected it with live Treasury and market data for real-time prices and offers.

Result:

• Delivered a fully integrated bond trading system that enhanced trading efficiency and connectivity with external market participants.

Challenge:Develop a CDS trading system for Asset-Backed and Mortgage-Backed Securities as part of a global initiative to consolidate CDS trading platforms.

Process:

• Designed and developed a multi-threaded pipeline to ensure a highly responsive UI with concurrent updates using C#, WinForms, MVC, and SQL Server.
• Collaborated with the global team to consolidate all CDS trading platforms.

Result:

• Delivered a responsive and efficient CDS trading system, significantly enhancing user experience and operational efficiency across consolidated platforms.